选取无人为干扰区、人为干扰区共55个土壤样品测定全氮磷钾含量,结合实测光谱的微分变换建立预测模型,并筛选出不同人为干扰下最优模型。结果表明:人类活动造成土壤全氮磷钾含量降低,数据集中程度和分布状态均发生变化。无人为干扰区以二阶微分建立的全氮、全钾含量预测模型和全磷以对数的倒数一阶微分建立的模型建模集r^2均超过0.9,检验集相对分析误差(RPD)均大于2.0,预测效果最佳。人为干扰区以倒数一阶微分建立的全氮、全磷含量预测模型不仅RPD>2.0,而且r^2> 0.92,也达到极好水平。入选RPD> 2预测模型的敏感波段中,无人为干扰区均位于可见光近红外波段,而人为干扰区则在近红外波段。这为今后提高全氮磷钾含量预测精度提供了新视角。
Total nitrogen,phosphorus and potassium in 55 soil samples from unmanned interference area and human interference area were determined. The prediction models were established based on the differential transformation of the measured spectrums,and the best models under different human interference were screened out. The results showed that the total amount of nitrogen,phosphorus and potassium in soil decreased due to human activities,and the concentration and distribution of data varied. In the unmanned interference area,the model sets r^2 of the total nitrogen and potassium prediction models established by second-order differential and the first-order derivative of the inverse of the logarithm of total phosphorus were more than 0.9,and the test sets RPD were over 2.0.These models have better prediction results.The total nitrogen and phosphorus prediction models established by the reciprocal first-order differential in the human interference area not only achieve an excellent level of RPD>2.0 but also r^2>0.92. In the sensitive bands selected for the RPD>2 prediction models,the unmanned interference area were located in the visible near-infrared band while the human interference area were in the near infrared band. This provides a new perspective for improving the prediction accuracy of total nitrogen,phosphorus and potassium in the future.
Journal of Desert Research